GHSA-wrc3-8c98-283w: 
Python vulnerability analysis and mitigation

Summary

An Improper Restriction of XML External Entity Reference ('XXE', CWE-611) vulnerability in AI RAG document ingestion parsers (lxml.etree.fromstring(resolve_entities=True) / xml.dom.pulldom) allows remote attackers to disclose local host files (file:///etc/passwd, .env, AWS credentials) or trigger SSRF by uploading a crafted XML, SVG, or DOCX document into the RAG knowledge base.

Proof of Concept (PoC)

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE foo [ <!ENTITY xxe SYSTEM "file:///etc/passwd"> ]>
<document><content>&xxe;</content></document>

Remediation

Disable DTDs, external entities, and network access in XML parsers (resolve_entities=False, no_network=True, load_dtd=False) or use defusedxml.